组合预测方法在玉米施肥预测中的应用.pdfVIP

吉林农业大学学报 2010, 32( 3) : 312~ 315 http :/ / xuebao.j lau. e u. cn Journal of Jilin Agricultural University Email :j ln xb@ vip . sina. com 苏恒强, 朱春娆, 温长吉 吉林农业大学信息技术学院, 长春130118 : 为提高玉米施肥量预测的精度, 利用拉格朗日乘数法, 于预测误差平方和最小这一目标, 对肥料效 应函数神经网络施肥预测方法这2 种单一施肥量预测模型进行加权组合, 建立了玉米施肥组合预测模型 预测结果显示: 肥料效应函数神经网络施肥预测方法和组合预测模型的预测误差平方和分别为2 78940, 65379, 42172, 说明玉米施肥组合预测模型优于单一施肥预测模型; 采用组合预测方法对玉米施肥量进行预 测, 能够显著提高预测精度 : 组合预测; 玉米施肥; 预测模型; 拉格朗日乘数法 : TP18; S513062 : A : 20 10) Application of Combination Forecasting Method in Corn Fertilization Forecasting SU Hengqiang, ZHU Chunrao, WEN Changji College of Inf ormation Technology, Jilin Agricultural University, Changchun 130118, China Abstract: A combination forecastin model was applied to forecast fertilizin amount for corn to improve the precision of the prediction. In order to achieve the least sum of square error, the wei ht of fertilizer effect function method and neural network forecastin model in combination forecastin model was objec tively calculated with La ran e multiplier method, and a combination forecastin model for corn fertilizin amount was constructed. The results show that the error sum of squares of fertilizer effect function method, neural network forecastin model and combination forecastin model is 2 78940, 65379 and 42172 , respectively. The result accounts for the fact that combination forecastin model is superior to sin le forecastin methods, and adoptin combination forecastin model in corn fertilization can obviously improve forecastin accuracy. Key words: combination forecastin ; corn fertilization; forecastin model; La ran e multiplier method ,